messersc
commented
4 years ago Hi,
not sure if that's useful for you after this much time. You are right, your loss curves look very suspicious. I can run your notebook fine, looks like some newer version broke something, not sure what though. Might be that python 3.7 needs to be downgraded.
Find below the conda env that I used for maui:
# packages in environment at /vol/home-vol3/wbi/messersc/miniconda3/envs/maui:
#
# Name Version Build Channel
_libgcc_mutex 0.1 main
absl-py 0.8.1 pypi_0 pypi
astor 0.8.1 pypi_0 pypi
attrs 19.3.0 pypi_0 pypi
backcall 0.1.0 pypi_0 pypi
bleach 3.1.0 pypi_0 pypi
ca-certificates 2019.11.27 0
cachetools 3.1.1 pypi_0 pypi
certifi 2019.11.28 py36_0
chardet 3.0.4 pypi_0 pypi
cycler 0.10.0 pypi_0 pypi
decorator 4.4.1 pypi_0 pypi
defusedxml 0.6.0 pypi_0 pypi
entrypoints 0.3 pypi_0 pypi
gast 0.2.2 pypi_0 pypi
google-auth 1.8.1 pypi_0 pypi
google-auth-oauthlib 0.4.1 pypi_0 pypi
google-pasta 0.1.8 pypi_0 pypi
grpcio 1.25.0 pypi_0 pypi
h5py 2.10.0 pypi_0 pypi
idna 2.8 pypi_0 pypi
importlib-metadata 1.3.0 pypi_0 pypi
ipykernel 5.1.3 pypi_0 pypi
ipyparallel 6.2.4 pypi_0 pypi
ipython 7.10.1 pypi_0 pypi
ipython-genutils 0.2.0 pypi_0 pypi
ipywidgets 7.5.1 pypi_0 pypi
jedi 0.15.1 pypi_0 pypi
jinja2 2.10.3 pypi_0 pypi
joblib 0.14.1 pypi_0 pypi
jsonschema 3.2.0 pypi_0 pypi
jupyter 1.0.0 pypi_0 pypi
jupyter-client 5.3.4 pypi_0 pypi
jupyter-console 6.0.0 pypi_0 pypi
jupyter-core 4.6.1 pypi_0 pypi
keras 2.2.5 pypi_0 pypi
keras-applications 1.0.8 pypi_0 pypi
keras-preprocessing 1.1.0 pypi_0 pypi
kiwisolver 1.1.0 pypi_0 pypi
libedit 3.1.20181209 hc058e9b_0
libffi 3.2.1 hd88cf55_4
libgcc-ng 9.1.0 hdf63c60_0
libstdcxx-ng 9.1.0 hdf63c60_0
markdown 3.1.1 pypi_0 pypi
markupsafe 1.1.1 pypi_0 pypi
matplotlib 3.1.2 pypi_0 pypi
maui-tools 0.1.7 dev_0 <develop>
mistune 0.8.4 pypi_0 pypi
more-itertools 8.0.2 pypi_0 pypi
nbconvert 5.6.1 pypi_0 pypi
nbformat 4.4.0 pypi_0 pypi
ncurses 6.1 he6710b0_1
notebook 6.0.2 pypi_0 pypi
numpy 1.17.4 pypi_0 pypi
oauthlib 3.1.0 pypi_0 pypi
openssl 1.1.1d h7b6447c_3
opt-einsum 3.1.0 pypi_0 pypi
packaging 19.2 pypi_0 pypi
pandas 0.25.3 pypi_0 pypi
pandocfilters 1.4.2 pypi_0 pypi
parso 0.5.1 pypi_0 pypi
pexpect 4.7.0 pypi_0 pypi
pickleshare 0.7.5 pypi_0 pypi
pip 19.3.1 py36_0
pluggy 0.13.1 pypi_0 pypi
prometheus-client 0.7.1 pypi_0 pypi
prompt-toolkit 3.0.2 pypi_0 pypi
protobuf 3.11.1 pypi_0 pypi
ptyprocess 0.6.0 pypi_0 pypi
py 1.8.0 pypi_0 pypi
pyasn1 0.4.8 pypi_0 pypi
pyasn1-modules 0.2.7 pypi_0 pypi
pygments 2.5.2 pypi_0 pypi
pyparsing 2.4.5 pypi_0 pypi
pyrsistent 0.15.6 pypi_0 pypi
pytest 5.3.1 pypi_0 pypi
python 3.6.9 h265db76_0
python-dateutil 2.8.1 pypi_0 pypi
pytz 2019.3 pypi_0 pypi
pyyaml 5.2 pypi_0 pypi
pyzmq 18.1.1 pypi_0 pypi
qtconsole 4.6.0 pypi_0 pypi
readline 7.0 h7b6447c_5
requests 2.22.0 pypi_0 pypi
requests-oauthlib 1.3.0 pypi_0 pypi
rsa 4.0 pypi_0 pypi
scikit-learn 0.22 pypi_0 pypi
scipy 1.2.1 pypi_0 pypi
seaborn 0.9.0 pypi_0 pypi
send2trash 1.5.0 pypi_0 pypi
setuptools 42.0.2 py36_0
six 1.13.0 pypi_0 pypi
sqlite 3.30.1 h7b6447c_0
tensorboard 1.14.0 pypi_0 pypi
tensorflow 1.14.0 pypi_0 pypi
tensorflow-estimator 1.14.0 pypi_0 pypi
termcolor 1.1.0 pypi_0 pypi
terminado 0.8.3 pypi_0 pypi
testpath 0.4.4 pypi_0 pypi
tk 8.6.8 hbc83047_0
tornado 6.0.3 pypi_0 pypi
traitlets 4.3.3 pypi_0 pypi
urllib3 1.25.7 pypi_0 pypi
wcwidth 0.1.7 dev_0 <develop>
webencodings 0.5.1 pypi_0 pypi
werkzeug 0.16.0 pypi_0 pypi
wheel 0.33.6 py36_0
widgetsnbextension 3.5.1 pypi_0 pypi
wrapt 1.11.2 pypi_0 pypi
xz 5.2.4 h14c3975_4
zipp 0.6.0 pypi_0 pypi
zlib 1.2.11 h7b6447c_3
maui_vignette.zip
I like maui and explore the possibility of VAEs in a cancer subtyping project, but I am having a hard time to reproduce your vignette results. The issue arises at plotting the losses, I pretty much have no recorded loss, and from then on everything goes south, ROC curves are deflated etc. I wonder if it has anything to do with me using Keras 2.3? I got this user warning:
It almost looks as the fit function doesn't update weights from one epoch to another. Have you encountered such error? I attach my vignette, went a bit past the loss plot and then I stopped testing, but at teh end you can see my pip freeze.